analysis of particulate matter emissions from light-duty gasoline vehicles in kansas city

8/13/2019 Analysis of Particulate Matter Emissions From Light-Duty Gasoline Vehicles in Kansas City

1/82

Analysis of Particulate Matter Emissionsfrom Light-Duty Gasoline Vehicles in

Kansas City


2/82

Analysis of Particulate Matter Emissionsfrom Light-Duty Gasoline Vehicles in

Kansas City

Edward Nam

Carl Fulper

James Warila

Joseph Somers

Harvey Michaels

Richard Baldauf*

Richard Rykowski

Carl Scarbro

Assessment and Standards Division

Office of Transportation and Air Quality

And

*Air Pollution Prevention and Control Division

Office of Research and Development

U. S. Environmental Protection Agency

NOTICE

This technical report does not necessarily represent final EPA decisions or

positions. It is intended to present technical analysis of issues using data

that are currently available. The purpose in the release of such reports is to

facilitate the exchange of technical information and to inform the public of

technical developments.

EPA420-R-08-010

April 2008


3/82

Table of Contents

1 Executive Summary................................................................................................................ 12 Introduction and Background ................................................................................................. 3

2.1 Emission and Fuel Regulations....................................................................................... 4

2.2 Causes of Gasoline PM Emissions ................................................................................. 63 Test Program and Report Goals.............................................................................................. 8

3.1 Test Program Goals......................................................................................................... 83.2 EPA Analysis Report Goals............................................................................................ 9

4 Site Selection .......................................................................................................................... 95 Vehicle Recruitment Process and Procedures....................................................................... 116 Vehicle Testing ..................................................................................................................... 117 Aggregate PM Results .......................................................................................................... 15

7.1 Random Stratified Vehicle Recruitment Results .......................................................... 157.2 Emission Results........................................................................................................... 227.3 Comparison with past studies ....................................................................................... 26

7.4 Temperature Effects on Composite Data...................................................................... 287.5 PM Emissions Trends from Composite Data................................................................ 33

8 Bag Analysis of Kansas City Data........................................................................................ 408.1 Temperature Analysis of Bag Data............................................................................... 438.2 Model Year Trends by Bag........................................................................................... 468.3 PM to HC Ratios........................................................................................................... 488.4 Elemental (Black) to Organic Carbon (EC/OC) Ratios................................................ 51

9 An Examination of the Representativeness of the Emissions Data ...................................... 5510 Preliminary Inventory Results for Calendar Year 2002 ................................................... 6111 Conclusions....................................................................................................................... 6512 Acknowledgments............................................................................................................. 67

12 References......................................................................................................................... 6813 Appendix. PM2.5 Base Emission Rates for Passenger Cars and Trucks for calendar year2005. 76

i


4/82

List of Tables

Table 1 Demographic Comparison of MARC Survey Cohort to Census 2000........................... 17Table 2 Comparison of MARC Survey Cohort to Census 2000 in terms of County of Residence

Table 7. Composite Gravimetric Particulate Matter Results for Repeat Back-to-Back Tests

Table 11. Exponential emissions dependence on temperature (P


5/82


6/82

Figure 35. Temperature adjusted natural logarithm of cold start PM as a function of model yearbin ................................................................................................................................................. 48

Figure 45. HC and PM distributions in Kansas City for model years 1996+ with normal curves

Figure 46. A comparison of the HC emissions (by model year group and vehicle type) between

Figure 47. Comparisons of the CO and NOx emissions (by model year group and vehicle type)

Figure 48. A comparison of the log-normal distributions of hot running HC in Kansas City vs

Figure 36. The ratio of PM to HC as a function of model year. ................................................... 48Figure 37. Distribution of PM/HC ratios. ..................................................................................... 49Figure 38. The ratio of cold start PM to HC as a function of model year. ................................... 49

Figure 39. The ratio of bag 2 PM to HC as a function of model year. ......................................... 50Figure 40. Plots of bag 1 vs bag 2 emissions for HC and PM respectively................................. 50Figure 41. Comparison of Photoacoustic to TOR EC measurements on a logarithmic scale. .... 52Figure 42. Elemental Carbon to Total PM ratio as a function of test temperature....................... 53Figure 43. Elemental Carbon to Total PM ratio as a function of vehicle model year. ................ 54Figure 44. EC/PM ratio as a function of vehicle inertial weight. ................................................ 54

overlaid. ........................................................................................................................................ 56

Kansas City and Arizona (with a 1:1 line included)..................................................................... 58

between Kansas City and Arizona (with 1:1 lines included)........................................................ 59

Arizona.......................................................................................................................................... 60Figure 49. log-normal distributions of hot running HC from KC and AZ on a linear scale........ 60Figure 50. MOVES-NMIM Comparison: National Average and a Cold State. ........................... 63Figure 51. MOVES-NMIM Comparison: Two Southern States. ............................................... 64

iv


7/82

1 Executive Summary

This report summarizes and analyzes the particulate matter data from the Kansas City PMCharacterization Study conducted in 2004 and 2005. It supplements the Eastern Research Group

contractor report to EPA (2008), which describes the details of the study and provides somepreliminary results. The Kansas City PM Characterization Study had many different goals andmeasurements thus producing a large array of data. This report analyzes only certain crosssections of this large data set within the scope of the subsequent modeling that is required toimplement emission rates into an inventory model.

It should be first noted that PM is a dynamic pollutant that is constantly being influenced by itsenvironment therefore its formation is constantly changing both in the exhaust stream and in theambient air. Our tests are a snapshot using specific measurements under specific laboratory andthermodynamic conditions. Real-world PM may differ significantly.

The first part of this report contains an evaluation of whether the data collected in the KansasCity Study is properly representative of the light duty vehicle fleet. Based on our review of therandom sample of vehicle used, we conclude that the program was largely successful in properlycapturing vehicles by household size, age, residence type, and household income that isdemographically representative of the Kansas City metropolitan area.

We also evaluated whether high emitters are properly represented in the vehicle sample. Due tolack of data from other sources about fractions of PM high emitters in the vehicle fleet, welooked at high emitter rates for carbon monoxide and hydrocarbons among the vehicles in theKansas City study compared to other studies based on remote sensing and I/M data. Weconcluded that CO and HC high emitter rates for older vehicles from the Kansas City Study were

comparable with other sources; however there is less certainty whether the program captured thedirtier HC emissions from newer vehicles. We believe that the evidence implies that a properrepresentation of high emitters were captured in this study, however further analysis is requiredin order to prove this conclusively.

The remaining parts of the report describe in detail how the Kansas City data were analyzed andthe results of that analysis. Key results are summarized in the following paragraphs.

For PM, 50% of the emissions came from 13% of the vehicles. It was also found that lighttrucks had slightly higher PM emissions than cars. These results are generally consistent withpast studies.

The emission trends show a clear drop in emissions levels (for PM as well as HC, CO, NOx)with later model year vehicles. However, we have not yet determined whether the drop is due totechnology changes (compliance with tighter standards) or whether it reflects varying levels ofvehicle deterioration. More than likely, it is a combination of both of these factors, butquantifying this distinction will be reserved for a future publication.

1


8/82

Elemental or black carbon accounts for roughly 20% of the PM emissions, with the organiccarbon accounting for the rest. It was found that elemental carbon roughly doubles during startscompared to hot running operation. However the fractions were not found to depend heavily onmodel year or temperature. These results may be important to studies that attempt to relate PMemission inventories to ambient PM concentrations.

The testing program measured a number of vehicles under different temperature conditions(summer and winter). The study results indicate that PM increases exponentially as temperaturedecreases so that for every 20 F drop, PM doubles. This effect is more pronounced for coldstarts.

Applying the KC data in a draft version of EPAs MOVES model results in an estimated averagenationwide increase of light duty gasoline PM emissions of about 1.6 times compared toMOBILE6.2. Emissions are generally higher than MOBILE6.2 in winter months and lower thanMOBILE6.2 in summer months. Overall annual emissions are expected to be significantlyhigher in areas with colder winters, even while summer emissions in those areas may be lower

compared to MOBILE6.2. However, these comparisons do not fully account for all localconditions which may have an impact on a local inventory analysis. Because we have not yetcompleted our analysis of the relative impacts of deterioration and technology, we cannot yetpredict how PM emissions in MOVES and MOBILE6.2 will compare for future years. Inaddition, PM from light-duty gasoline sources only form a fraction of the overall PM inventory,where stationary, non-road, diesel, road dust, wood-burning, and many other sources (natural andman-made) also play a significant role. However, even for light-duty gasoline PM, there is muchwork to be done before a final estimate of inventory impacts can be determined.

In the future, EPA will continue to investigate factors that contribute to or reduce the formationof PM. EPA has also observed the variability of measurements (even for back-to-back tests) andwill continue to explore testing methodologies and procedures that may contribute to the non-repeatability of some measurements. It is also important to resolve the differences betweenKansas City and the more numerous inspection and maintenance data. In the future, it would beimportant to examine trends in the speciated hydrocarbons and organic PM from the standpointof toxic emissions and also to quantify the PM emissions attributable to oil consumption. This islikely to expand the scientific understanding of PM formation. For modeling purposes, it isimportant to understand the modal or load-based behavior of PM as well as determine therelative impacts of technology vs deterioration. Resolution of these topics will help us to updateEPAs inventory model, MOVES, in order to better generate inventories from the past and intothe future.

2


9/82

2 Introduction and Background

In 1998, the Coordinating Research Council conducted major studies on particulate emissionsfrom in-use vehicles in its Project E-24. This work was done in San Antonio, TX (by Southwest

Research Institute), Denver, CO (by a variety of groups including General Motors, the ColoradoDepartment of Public Health and Environment, Colorado State University, EPA, and the CleanAir Vehicle Technology Center), and in California (by the Center for Environmental Researchand Technology of the University of California at Riverside). This work, discussed elsewhere inthis paper, involved testing of several hundred vehicles and was designed to obtain PM emissiondata on vehicles of different model years in different locations. This work, as had other projects,showed the presence of high-emitting vehicles which have a substantial impact on overall PMemissions from gasoline-fueled vehicles. This work was extremely valuable in providing initialdata for emission factors on gasoline vehicles. Also, numerous source apportionment studies(discussed later in this paper) showed the large contribution to overall ambient PM fromgasoline-fuel vehicles. However, none of these studies were designed to determine the

frequency of high-emitting vehicles in the overall fleet or to accurately predict the emissionsfrom the overall fleet.

A major limitation in previous emissions testing studies has been the way vehicles have beenrecruited. Most studies have not incorporated random sampling in the study design due to thehigh non-participation rate and the high costs associated with generating and implementing arandom sampling plan. Therefore, few studies, and no studies evaluating light-duty PMemissions, can be used to represent the actual distribution of vehicle emissions in a largepopulation. Most test programs select the first vehicle that meets the test program vehiclespecifications usually based on model year, manufacturer, make, engine family or odometer.Gathering emission data from vehicles this way will provide what that particular vehicle is

emitting in the laboratory but will not tell you whether that vehicle is representative of an entiregroup of similar vehicles. Therefore, modelers are always trying to determine how to applythese emission data to represent the vehicle fleet over various geographic scales. The NationalResearch Councils report on modeling mobile-source emissions released in 2000 stated as partof their recommendations that EPA should:

Develop a program to enable more accurate determination of in-use emissions; Begin a substantial research effort to characterize high exhaust emitting vehicles; Update their models with the best available data on PM emissions, and Incorporate estimates of mobile-source toxic emissions into our models.

EPAs staff started developing and proposing a test plan in 2001 to foster interest from potentialparties for this type of a test program. Through this effort, EPA was able to develop aconsortium of sponsors by early 2003 that included: the Coordinating Research Council (CRC),the U.S. Department of Energys (DOE) National Renewable Energy Laboratory (NREL), theU.S. Department of Transportation (DOT) Federal Highway Administration (FHWA), and theState and Territorial Air Pollution Program Administrators/Association of Local Air PollutionControl Officials (STAPPA/ALAPCO) through EPAs Emission Inventory ImprovementProgram (EIIP). EPA also established a cooperative research and development agreement

3


10/82

(CRADA) with the CRC that allowed for their sponsorship of this test program and providedEPA with technical expertise. An advisory committee, consisting of most of the sponsors, wasestablished as an oversight committee for the test program. This committee advised EPAs staffon the proper testing methodologies, procedures and assisted in resolving any important issuesthat arose before, during and after the test program. EPA was responsible for managing the

testing contract, the contractor and making the final technical decisions on how the test programwould be conducted. After a competitive solicitation process and evaluation, EPA awarded thetesting contract to Eastern Research Group (ERG).

Some earlier test programs conducted by EPAs Office of Research and Development and othersindicated that temperature might influence the amount of PM emitted from light-duty vehicles,especially at colder temperatures (Stump et al., 2002; Cadle et al., 1999). To help address thesetemperature concerns, EPA split the study into two equal rounds of vehicle testing. One round ofvehicles was tested during the summer months and a second round of vehicles in the wintermonths. This allowed for the widest temperature profile. To evaluate trends between the rounds,41 vehicles from the summer test program also underwent testing in the winter phase. In

addition, a reference vehicle with well-characterized emissions was tested weekly throughout theprogram. The study also conducted detailed gaseous and PM speciation, including toxics, inexhaust emissions in approximately 25 vehicles each round.

Finally, all the data gathered under this program is undergoing data validation and is beinganalyzed by EPA to help meet some of the National Research Councils recommendations indeveloping better modeling tools for mobile-source emissions. All this data will be stored in theEPA OTAQs Mobile Source Observation Database (MSOD) and made available to the generalpublic for their scientific review. Information pertaining to the specific study design and qualitymanagement plan can be found in the supporting documentation (EPA, 2008).

2.1 Emission and Fuel Regulations

The Clean Air Act of 1970 gave EPA the broad authority to regulate motor vehicle pollution, andthe Agency has implemented multiple emission control policies to reduce emissions frompassenger cars and the light trucks. Efforts by government and industry since 1970 have greatlyreduced typical vehicle emissions. EPA has issued many successful control programs, theNational Low-Emission Vehicle (NLEV), Reformulated Gasoline (RFG) and Tier 2 vehicle andgasoline sulfur standards are important recent examples that will continue to help reduce car andlight-duty truck emissions into the near future. In that same period of time, however, the numberof vehicles and the distance driven have steadily increased. This increase in travel by passengercars and light trucks will continue to make motor vehicles significant contributors to air pollutioninventories well in the future.

Exhaust emissions of particulate matter from gasoline powered motor vehicles and dieselpowered vehicles have changed significantly over the past 25 years (Sawyer and Johnson, 1995;Cadle et al., 1999). These changes have resulted from reformulation of fuels especially theremoval of lead additives, the wide application of exhaust gas treatment in gasoline-poweredmotor vehicles, and changes in engine design and operation. Particularly, as emission standards

4


11/82

reduced exhaust hydrocarbons with the introduction of catalysts in 1975, the organic componentof exhaust PM also decreased. Lead, which was the major PM component in gasoline vehicleexhaust, was virtually eliminated with the introduction of unleaded gasoline mandated for the1975 model year vehicles and the later phase-out of lead in all motor vehicle gasoline.

For some time, it has been well-known that in-use (gas-phase) emissions of gasoline-poweredmotor vehicles can be significantly higher than the standards to which they were certified whennew. Age or odometer-related deterioration, engine or fuel system malfunction, broken (orremoved) catalysts, and poor owner maintenance can result in higher emissions. As a result,prior emission studies have indicated that emissions from similar vehicles can span severalorders of magnitude (Hildemann et al., 1991; Cadle et al., 1997; Sagebiel et al., 1997; Yanowitzet al., 2000). Because of evolving tailpipe emission standards, along with the wide variability ofemissions between vehicles of the same class, well-defined average emissions profiles for themajor classes of motor vehicles have been difficult to establish.

EPA has regulated HC, CO, and NOx exhaust emissions from gasoline vehicles since 1968.

These regulations have become increasingly stringent with time including those for model year1975 which resulted in introduction of catalysts, 1981 when the 3-way catalyst was widely used,and in later years such as for the 1994 model year [check year] when the Tier 1 standards wereintroduced. In contrast to EPAs strict regulations on diesel smoke (effective in the 1970 modelyear) and diesel PM (effective with the 1988 model year), EPA did not regulate PM emissionsfrom gasoline vehicles until relatively recently. The first PM standards were part of the Tier 1regulations, which phased in starting with the 1994 model year, through 1996. These standardswere designed to provide more of an upper limit on PM emissions rather than to effect actualreductions. EPA further regulated gasoline PM as part of the Tier 2 regulations [CFR 65, 2000]ranging from 0.00 g/mile for Tier 2 bin 1 (for zero-emission vehicles) or 0.01 g/mile for Tier 2bins 2 to6) to 0.08 g/mile for Tier 2 bin 10. These regulations were phased in from the 2004 to2006 model year. Since new model year gasoline vehicles typically have lower PM emissions,emission testing is not generally done on these vehicles in the certification process.

The majority of exhaust PM emitted by catalyst-equipped motor vehicles is in the PM2.5sizerange (particulate matter mass with aerodynamic size of 2.5um or less, typically collected on afilter). Kleeman et al., (2000) have shown that gasoline and diesel fueled vehicles produceparticles that are mostly less than 2.0 m in diameter. Cadle et al., (1999) found that 91% of PMemitted by in-use gasoline vehicles in the Denver area was in the PM2.5size range, whichincreased to 97% for smokers (i.e., light-duty vehicles with visible smoke emitted from theirtailpipes). Durbin et al.,(1999) found that 92% of the PM was smaller than 2.5 m for smokers.The mass median diameter of the PM emitted by the gasoline vehicles sampled by Cadle et al.,(1999) was about 0.12 m, which increased to 0.18 m for smokers.

Gasoline PM consists mostly of carbonaceous components including elemental carbon as well asthose derived from organic constituents, generally higher molecular weight hydrocarbons.Gasoline PM consists largely of the higher molecular weight hydrocarbon/organic constituentswhich comprise approximately 60-99% of the total PM mass. Elemental carbon comprises arelatively small fraction of the PM mass. A figure of 26% of gasoline PM being element carbonhas been found in the emission testing done for the Northern Front Range Air Quality Study

5


12/82

(Watson et al, 1998) . Vehicles that burn oil (frequently older vehicles) have a higher amount ofPM mass and, generally, a higher fraction of high molecular weight organics (certainly thoseassociated with oil combustion) in the PM. Vehicles running rich (e.g. under enrichment orduring cold temperature starting) can emit higher levels of elemental carbon (black soot).

Gasoline PM also contains a small amount of sulfate, generally in the form of sulfuric acid, fromoxidation of the SO2formed in combustion. A fraction of the SO2(roughly 10%)is oxidizedover the catalyst to form sulfuric acid/sulfate. Non-catalyst vehicles have roughly 1% of the SO2in vehicle exhaust further oxidized to sulfuric acid/sulfates. The fraction of gasoline PM that issulfate varies but has been prior to 2006 generally no more than 5% of total PM mass making it arelatively small constituent. About 100 vehicles were tested in an EPA baseline study on sulfateemissions which is a major basis for PM data on sulfates (Somers et al 1977). However,effective in 2006, EPA Tier 2 regulations resulted in the reduction of gasoline sulfur content to30 ppm from an average of roughly 300 ppm. This reduction in gasoline sulfur content shouldresult in gasoline PM containing much smaller amounts of sulfates.

2.2 Causes of Gasoline PM Emissions

In this section, we briefly summarize factors that contribute to gasoline PM in the vehicle fleet.Where appropriate, we will also compare to the mechanisms of hydrocarbon (HC) formation,since parallels are often drawn in the literature. Particulate matter is formed from gasoline-fueled engines from incomplete fuel and oil combustion. The amount of oil consumed incombustion and its contribution to PM varies greatly from vehicle to vehicle. There arenumerous distinct technologies used in vehicles, which are in various states of repair or disrepairwhich also affect PM emissions. Even brand new vehicles emit PM from combustion but at verylow levels. While a complete description of what causes PM emissions and the mechanisms

behind it is beyond the scope of this report, there are many aspects of the science that are still notwell understood.

Simply put, particulate matter primarily forms during combustion (and afterwards) when carbon-containing molecules condense into solid form. This PM is generally higher molecular weighthydrocarbon compounds, some of which originate in the fuel/oil and some of which are formedin combustion. Unlike diesel engines, elemental (molecular) carbon or soot is not very prevalentwith gasoline engines but does form in larger quantities under relatively rich air:fuel ratios. Theamount of elemental carbon in PM varies from vehicle to vehicle (and, even for a given vehicle,varies depending on operating conditions and state of repair). There are also other compounds inthe fuel or engine oil such as trace levels of sulfur and phosphorus which, in combustion, formsulfates and phosphates, both of which are PM. The sulfur level in gasoline is now very lowalmost eliminating sulfate formation from gasoline sulfur content but motor oil containssignificant sulfur (and phosphorus) compounds. Also, trace metal constituents in gasoline andoil forms PM in the combustion process as metallic oxides, sulfates, nitrates, or othercompounds. Attrition products from the catalyst substrate and trace amounts of noble metals canalso be emitted as PM. The catalyst attrition products are mechanically generated and areusually coarse particles (>2.5um). Exhaust PM as formed in the engine is generally very small in

6


13/82

size (possibly much of it is nuclei mode PM in the 0.05 micron or smaller size range). In theexhaust system including the muffler, some of the PM agglomerates and grows in size.

There is a wide assortment of technologies in vehicles that can affect PM formation. Thesetechnologies were mainly developed to control HC, CO and NOx emissions, but most have the

side benefit of also reducing PM since reducing exhaust HC generally also reduces exhaust PMalthough not to the same extent. Older engines from the 1980s and earlier that deliver fuelthrough a carburetor typically have poorer fuel droplet quality, as well as poorer control of fuelair stoichiometry. These older vehicles are expected to produce more PM (on average) then theirfuel injected engines that followed generally in the late 1980s and early 1990s. Among fuelinjected engines, throttle body fuel injection (TBI) used in earlier engines with fuel injectiontypically has poorer fuel atomization quality and air:fuel ratio control than the port fuel injection(PFI) technology that supplanted it; thus, one might expect older model year fuel injectionvehicles to have higher PM emissions (on average) than newer ones. Somewhat before thewidespread use of fuel injection, closed loop control systems were developed along with oxygensensors to improve the stoichiometric chemistry of combustion of catalytic conversion. These

closed loop controls improved combustion as well as the effectiveness of the after-treatmentsystem. The after-treatment system on most vehicles consists of a 3-way catalyst. The 3-waycatalyst was designed for simultaneous control of hydrocarbons, carbon monoxide, and nitrogenoxides. Vehicles with 3-way catalysts would meet more stringent hydrocarbon and carbonmonoxide emission standards while also meeting the first stringent nitrogen oxide standard. Inoxidizing hydrocarbons, these systems are resulted in additional PM control. These systemswere utilized on almost all gasoline-fueled vehicles beginning in the 1981 model year. On somemodel-year vehicles in the 1980s and a few more recently, a secondary air injection system wasadded between the engine and oxidation portion catalyst in order to supplement air to theoxidation reactions on the catalyst. These systems also helped oxidize PM also (though probablynot to the extent that it oxidizes CO or HC). The deterioration of these technologies may affectPM and HC quite differently.

The amount of PM is very sensitive to the amount of fuel in combustion as well as the air:fuelratio. Over-fueled mixtures results in higher PM formation and, in some cases, also in excesssoot formation. Over-fueling can occur under several different conditions. During cold start,engines are often run rich in order to provide sufficient burnable fuel (i.e. light ends that vaporizeat colder temperatures) to start combustion when the cylinder walls are still cold. When highacceleration rates or loads are encountered (such as in a wide open throttle event), an extraamount of fuel is often injected, resulting for greater power or catalyst and componenttemperature protection. Emission control systems in the late 1990s are designed to limit thisenrichment. Finally, engines can run rich when a control sensor (e.g. oxygen, MAF, MAP, orcoolant sensors) or the fuel system fails.

In addition to fuel, lubricating oil can also get into the combustion chamber via severalpathways. Engine components, such as valves, valve seals, piston rings, and turbochargers canwear and deteriorate. During the intake stroke, the negative pressures (engine intake vacuum)can pull oil through the gaps left by these worn parts. In all gasoline automotive engines, thecrankcase (where the oil splashes onto the engine components) is vented back into thecombustion chamber through the intake manifold. This is known as Positive Crankcase

7


14/82

Ventilation (PCV), and is required in order to remove and burn the excess hydrocarbons in thehot crankcase. Unfortunately, it can also introduce PM precursors and oil into the enginecombustion chamber. Because of the relatively small amount of oil consumption compared tothe volume of gasoline burned in a vehicle, HC from oil is also small. However, organic PMfrom oil consumption can be quite significant because oil is a high molecular weight

hydrocarbon, and more likely to be in uncombusted droplets. Therefore as vehicles get older,those that consume more oil, will probably have very different HC emissions behavior than PM,compared to when it was new. However, oil consumption can "poison" the catalyst materialreducing the effectiveness of the catalyst at oxidizing HC.

Some of these PM forming mechanisms clearly affect HC emissions. So a control technology ora deterioration path for HC may or may not have the same effect on PM depending on thesource. It is also likely that the processes that cause high PM may not be the same processes thatcause organic PM. Some of the mechanisms also form visible smoke. Smoke takes on a varietyof characteristics depending on the source, and can be due to oil consumption or overfueling.The smoke is visible because of the relative size of the particles compared to the light

wavelengths that are scattered. Visible smoke is however not a necessity for high PM emissions.

Finally, the fuel itself may have properties that exacerbate PM formation. These are sensitive tothe concentrations of the following: lead, sulfur, aromatics, and impurities. With the lower levelsof lead and sulfur in fuels recently, the first two are probably less of a factor in the Kansas Cityprogram than aromatics would be.

3 Test Program and Report Goals

3.1 Test Program Goals

The program was designed to estimate average PM emissions for the fleet with attention tocharacterizing the contribution of high-emitting vehicles. A large body of previous work hasdemonstrated that a small fraction of motor vehicles emits a disproportionate fraction ofparticulate emissions. These high-emitters can exhibit higher emissions rates under certainspecific operating or environmental conditions than do normal emitters. During the design ofthe study, we concluded that no reliable method of screening or identifying high-emittingvehicles short of actually measuring them was available. Thus, to use resources effectively, thegoal was to employ a sampling design to provide a context for understanding and interpreting theresults, including the frequency and contributions of both normal and high-emitting vehicles.

To achieve this goal, the approach adopted was to over-sample from portions of the fleet wherehigh-emitting vehicles were assumed to be most prevalent. This approach was implemented bysampling older vehicles in higher proportions than those vehicles existed in the fleet. Duringsubsequent analysis, EPA will investigate some of the variables and factors (mileage,maintenance, age, environmental conditions, engine design or other emission devicetechnologies, etc.) that might be an influence on these vehicles.

8


15/82

The study also had secondary objectives of advancing the development of analyticalmethods. One goal was to evaluate the capabilities of onboard portable emission measurementsystems (PEMS) in the collection of mass-based emissions data. A similar goal was to evaluatethe use of portable activity measurement system (PAMS) to monitor the activity of a vehicle(e.g. engine on/off, vehicle speed, date, time of day etc.) during normal driving. Usage and

testing of these instruments will allow EPA to advance the development of instrumentation andassociated protocols, and to demonstrate the utility of portable instrumentation in collection ofreal-world emissions and activity data. Unfortunately, PM was not measured on-road using thePEMS devices during this study, so discussion of PEMS results is beyond the scope of thispaper.

Another goal was to investigate the use and measurement of other continuous particulate mattermeasurement devices. Three devices, (DustTrak, DataRAM and Quartz Crystal Microbalance(QCM)), employing different measurement techniques were used during this test program.Another area of investigation was to advance the development of devices that can measure PMon a continuous basis versus traditional gravimetric measurements. Different versions of the

Quartz Crystal Microbalance (QCM) were evaluated in each round of vehicle testing. Some ofthe results from these measurements can be found in the EPA (and ERG report 2008).

3.2 EPA Analysis Report Goals

This report will discusses preliminary investigation and and analysis of different aspects of thePM data gathered during this program. In its review, EPA will begin by investigating if thestudy was able to achieve a random sample from the KCMA and if the non-respondents weresignificantly different from the positive respondents. This is followed by a discussion on EPAsanalysis pertaining to the distribution of PM emissions within the KCMA vehicle fleet. It will

include preliminary analysis on high-emitters or smokers and will compare results withprevious programs that have tried to identify such groups of vehicles. Third, the report willinvestigate the trends in the PM data. Trends include temperature effects on PM emissions aswell as correlations to model year, age, odometer, emission standards, other pollutants, etc., thatmight influence these findings. Finally, EPA will provide some analysis of PM modelingmethodologies that could be used to incorporate these findings and present current inventoryestimates. Future inventory estimates will be presented in another publication.

4 Site Selection

EPA started investigating medium to large metropolitan areas that would have a vehicle fleetrepresentative of typical PM emission for the light-duty vehicles nationwide. It was determinedearly in our evaluation that the area should not have an Inspection and Maintenance (I/M) orReformulated Gasoline (RFG) program. In an I/M program area, vehicles are regularly requiredto be tested and repaired to prevent their emissions from becoming excessive. Thus, I/Mprograms affect a vehicles deterioration rate by requiring repairs and maintenance that otherwisemight not normally be performed by the vehicle owner. This program changes the naturaldeterioration rate of vehicles. EPA also did not want metropolitan areas that had an RFG

9


16/82

program, since this program reduces the emissions from vehicles in the summer months. Thiswould increase the complexity in trying to remove this emission effect for vehicles in areas thatdo not have this program. By not selecting an area with these programs, EPA would beevaluating vehicles under natural deterioration rates and with standard fuel properties. EPAalso wanted to consider an area that would represent a typical metropolitan area that had a

developed transportation structure with moderate driving patterns, socioeconomic demographics,and a broad range of seasonal temperatures. After careful review, EPA selected the centrallylocated Kansas City (MO/KS) metropolitan area to conduct this landmark PM study. TheKansas City Metropolitan Area (KCMA) for this study consisted of the following counties:Johnson County, KS; Leavenworth County, KS; Wyandotte County, KS; Clay County, MO;Cass County, MO; Jackson County, MO; and Platte County, MO.

There were two secondary factors that influenced EPA in its selection of Kansas Citymetropolitan area (KCMA). The first factor was a prior transportation study that had beenconducted in the area recently that could potentially be used in support of vehicle recruitment.The second factor was the KCMA has had multiple emission test programs conducted usingremote sensing devices (RSD) which could be used as an additional tool to possibly help indetermining if the vehicles were recruited in a random basis.

Figure 1, Kansas City Metropolitan Area (KCMA)

10


17/82

5 Vehicle Recruitment Process and Procedures

EPA and its contractor (ERG) designed a vehicle recruitment process to characterize the averagePM emission factor from a representative sample of vehicles, with special attention paid to

recruitment of initial non-responders, who were thought to be potential drivers of higher-emittingvehicles. Vehicles were divided into eight vehicle type strata: two vehicle types (passenger carand trucks) and four model year age groups (pre-1981, 1981-1990, 1991 1995 and 1996 andnewer).

To achieve the random stratified vehicle recruitment goals, ERG used vehicles from atransportation study that was conducted by the Mid-America Regional Council (MARC) in 2004(Kansas City Regional Household Travel Survey Final Report, 2004). The MARC 2004Household study was a survey of KCMA regional households travel activities. This study wasused as the initial starting point for generating a stratified random sample. The surveyprocedures for establishing the cohort required the use of a random digit dial (RDD) telephone

survey of households (HHs) in the seven county KCMA area from which a sample of vehicleswere recruited for testing. The MARC data were compared to the latest U.S. Census data (2000)on key household characteristics including household size, vehicles, household workers,household income, residence type, household ownership and the types of vehicles owned.

Vehicles were randomly selected out of the cohort for testing. It became apparent that theMARC cohort had fewer than expected older vehicles available for recruitment and by the end ofround 1 testing, the contractor had exhausted its pool of older vehicles (pre-1981 and 1981-1990)to recruit for testing. In order to address this problem, EPA and our contractor acquired bothKansas and Missouri Vehicle Registration databases which provided a large pool of vehicles thatcan be sampled and recruited for testing. These databases were used to draw a stratified random

sample for recruiting the vehicles necessary to achieve the overall desired sampling strata targets.An incentive survey was also conducted to identify the appropriate levels of incentives necessaryto ensure sufficient regional vehicles would be available for the emissions test program. Furtheranalysis conducted on both the MARC and State vehicle registration database and the incentivesurvey can be found in the supporting documentation to this report (EPA, 2008).

6 Vehicle Testing

The KC study was conducted in three distinct phases: pilot testing, Round 1- summer testing

and Round 2 - winter testing. A pilot study was conducted in May 2004. The primary goals ofthe pilot study were to establish a temporary testing facility in the Kansas City area and tofinalize all testing methodologies, testing procedures and data handling procedures. Thecontractor and EPA staff also tested three EPA-provided correlation vehicles to compareEPAs National Vehicle and Fuel Emission Laboratory (NVFEL) dynamometer measurementswith those obtained using the EPAs Office of Research and Development (ORD) portableClayton dynamometer at the KC test facility. The details of the pilot study are discussedelsewhere (EPA, 2008- Appendix BB). The report identified procedural changes that were

11


18/82

incorporated into the Quality Management Plan (QMP) and Quality Assurance Project Plan(QAPP) that were approved before the start of Round 1 vehicle testing (EPA, 2008- AppendicesAA & II).

Summer testing (Round 1) started in July 2004. During this round, approximately 261 vehicleswere tested under summer conditions at the facility. The contractor had vehicles arrive at the testfacility one to two days before actual dynamometer emission testing would occur. Vehicleowners were scheduled at specified times to ensure quick and individual attention was given toeach owner. ERGs subcontractor, Nustats, was given the task of making sure that enoughvehicles were on at the test facility to meet the test program goals and that vehicles were beingrecruited on a random basis.

Upon arrival, each vehicle first received a unique identification code for documentation trackingpurposes, and was then inspected for test worthiness. Specific vehicle information, in the formof digital photographs, interview questionnaires, checklists, and hard copy data forms, wasrecorded for later input into the MSOD data table. A study was conducted by the contractor toevaluate what level of compensation was appropriate for the use of their vehicle and ownerstime. Some vehicles were picked up and/or delivered to the owners home or office to helpfacilitate the use of that vehicle in this study. The owner was also given the option to use one ofour loaner vehicles while their vehicle was tested.

The vehicle was inspected with the owner present to document the condition of the vehicle. Amore detailed inspection occurred later to ensure that the vehicle could safely be operated on theroad and dynamometer. If repairs were required, the vehicle owner was notified and theirpermission was obtained before repairs were performed. If the repairs could not be performedon-site, the vehicle was taken to a local repair shop. Records of the repair, along with a briefnarrative, were maintained. Typical repairs, for example, the replacement of brakes or part ofthe exhaust system were done for either safe operation of the vehicle or to allow for proper

emission testing. Any defects or deficiencies in the vehicle condition that might affect exhaustemissions where not repaired by the contractor. All vehicles were maintained in what is calledan as received or as is condition. During this inspection, a small sample of lubricating oiland, if possible, a fuel sample were taken and stored in a refrigerated unit for later possibleanalysis.

The next step in the process was to have the vehicle conditioned before being tested on thedynamometer. The main purpose for conditioning a vehicle was to make sure each vehicle wasoperated the same way for a certain period of time before being tested. This conditioningprocess should minimize the impact of the owners driving pattern and habits on the subsequentemission measurements which might have occurred via the engines adaptive learning capability.

The conditioning route developed by the contractor and approved by EPA was about 45 minuteslong and included high speed accelerations, driving at freeway speeds, and driving in stop-andgo traffic patterns through different roadway types: city, arterial and highway. The exact routedriven has been documented in the contractors report. During the conditioning process, aportable emission measurement system (PEMS) manufactured by Sensors Inc. was installed ontothe vehicle to monitor emissions. The incorporation of a PEMS device onto the vehicle duringits conditioning provided a couple of key quality assurance and control techniques. It firstprovided real world emission and activity data on vehicles that could be compared to

12


19/82

dynamometer data. It also allowed for the gathering of emission data on vehicles that could notbe tested on the dynamometer. These excluded vehicles might be too long, too wide, all-timefour-wheel drive, or the vehicles condition (engine) prevented the vehicle from performing thedrive cycle on the dynamometer. All of these vehicles would normally be excluded fromlaboratory testing. This program allowed the PEMS device to be installed on these vehicles and

have them driven on the conditioning route, therefore some of these data may be included intothe national emission profile analysis.

The PEMS unit used for the conditioning drive underwent a complete warm-up, zero and auditsequence to verify CO, CO2, NOx, and THC measurement accuracy. The contractor establishedprocedures including check lists to ensure proper installation and calibrations were performed asnecessary to bring the PEMS into proper calibration. The vehicle was driven on the conditioningroute described above. The PEMS unit was then uninstalled, data analyzed, and the vehicle wasleft inside the facility to soak overnight at ambient temperatures. The PEMS data was analyzedto determine the testing order of vehicles for the next day with the cleanest being tested first andthe dirtiest last. This was done to reduce any effect that a high emitting vehicle might have theon the emission testing equipment.

The following day, the vehicle was pushed onto the dynamometer, and secured. EPA usedORDs transportable Clayton Model CTE-50-0 twin-roll chassis dynamometer. Test inertia andhorsepower settings for the dynamometer were determined from EPA I/M lookup tables. APositive Displacement Pump-Constant Volume Sampling (PDP-CVS) system was used to diluteand transport the vehicle tailpipe exhaust to analyzers during the dynamometer test. Dilutiontunnel air was kept constant at 47C 5C to prevent loss of volatile PM components.Procedures for conditioning the tunnel and analytical equipment to minimize any release ofvolatile compounds was reviewed during the pilot study and checked daily and weekly duringboth rounds of tests. The PEMS unit was installed directly onto the vehicles tailpipe to monitorundiluted emissions, in tandem with the emissions measurements to be performed by the

dynamometer bench.

Vehicles were operated over the LA92 Unified Driving Cycle. The LA92 cycle consists of threephases or bags. Phase 1 (or bag 1) is a cold start that lasts the first 310 seconds. Coldstart is technically defined as an engine start after the vehicle has been soaking in atemperature controlled facility (typically ~72F) with the engine off. In the Kansas City study,the vehicles were soaked over night in ambient conditions. Phase 1 (310 seconds or 1.18 miles)is followed by a stabilized Phase 2 or hot running (311 1427 seconds or 8.63 miles). At theend of Phase 2, the engine is turned off and the vehicle is allowed to soak in the test facility forten minutes. At the end of the soak period, the vehicle is started again, and is driven on the samedriving schedule as Phase 1. This Phase 3 is called a hot start because the vehicle is started

when the engine and aftertreatment are still warm or hot. Criteria pollutants were measuredboth in continuous and bag modes. PM was gathered for each of the three Phases on 47 mmTeflon filters at 47C 2C.

13


20/82

70.0

50.0

60.0 Bag 1

Bag 2

40.0

speed(mph)

30.0

20.0

10.0

0.0

0 200 400 600 800

time

Figure 2. LA92 driving schedule with each of the bags identified.

1000 1200 1400

In addition to the regulated gas pollutants measured via the CVS, continuous measurements ofPM mass were taken using an EPA-supplied Booker Systems Model RPM-101 QCMmanufactured by Sensors Inc. and a Thermo-MIE Inc. DataRam 4000 Nephelometer. Anestimate of black carbon was measured continuously with a DRI photoacoustic instrument andintegrated samples were collected and analyzed by DRI for PM gravimetric mass, elements,elemental and organic carbon, ions, particulate and semi-volatile organic compounds, andvolatile organic air toxics. All sampling lines were heated and maintained at 47C 2C. The

samples were extracted from the dilution tunnel through a low particulate loss 2.5 m cutpointpre-classifier. Further, details and a schematic of the sampling instrumentation can be foundFigure 3 and in EPA, (2008).

At the conclusion of vehicle testing, the vehicle was disconnected from the PEMS anddynamometer sampling systems and removed from the dynamometer. The vehicle was released

to the owner after the tests were reviewed by the contractor to confirm a valid emission test hadoccurred. The CVS tunnel blowers were kept on and monitored by numerous analytical devicesto ensure that the tunnel had stabilized (No off-gasing of volatile organic compounds). Once,the tunnel had stabilized and proper analytical test procedures completed, the next vehicle waspushed onto the dynamometer and tested.

14


21/82

( 4 6 C )

D i l u t e d e x h a u s t

a t 4 6 C

Figure 3. CVS Sampling System Schematic

Winter testing (Round 2) started in January 2005. During this round, approximately 278vehicles were tested under winter conditions at the facility. Approximately 43 vehicles testedduring Round 1 were re-tested in Round 2 to estimate the effect of the ambient temperature onexhaust emissions. The same testing procedures were used except for a couple of small changes

as noted in the QAPP documentation in the KC EPA report (2008). The contractor provided toEPA their review of all data verification and validations that were performed throughout this testprogram. ERG recorded and reported all data and did not remove or eliminated any data. Alldata exceptions were noted within their report. Further details on these areas are documented inthe EPA report on the study (2008).

7 Aggregate PM Results

Many of the general data verification and validation results from the Kansas City program areincluded in the EPA report (2007). This section will report analysis that was either reviewed or

conducted by EPA and is divided into two main areas; analysis on recruiting a random stratifiedvehicle sample and on the KCMAs vehicle emissions.

7.1 Random Stratified Vehicle Recruitment Results

At the outset, the contractor analyzed the household cohort, to assess its representativeness withrespect to households within the KMCA. To achieve this goal, the cohort was compared to

15


22/82

results from Census 2000 on the basis of a series of demographic variables, including householdsize, income levels, home owners age and numbers of vehicles owned. Table 1 shows that thecohort distributions were generally similar to those from the Census. The largest differenceobserved with respect to residence type, in which the prevalence of vehicle owners living insingle-family dwellings was higher for the cohort than for Census, with the reverse being true for

other types. This difference is explainable in terms of the fact that the cohort was constructedusing random-digit dialing (RDD) and therefore related to patterns of telephone ownership. Ingeneral, the availability of listed telephone numbers is associated with length of tenure inresidence, which is in turn associated with home ownership and single-family dwellings.

It is also of interest to compare the demographic characteristics of vehicle owners participating inthe study to those for the cohort and the Census. Table 1 also presents demographic distributionsfor participants in Summer and Winter Phases. In these cases, some differences can be seenbetween the study participants and the cohort (and Census). One obvious difference is that onlyhouseholds owning vehicles could be included, explaining the absence of households withoutvehicles. With respect to numbers of household vehicles, households owning more vehicles

participated at higher rates than in the general population. With respect to income levels,households at the low and high ends of the income distribution participated at lower rates than inthe general population. In contrast, households in the middle income ranges participated atslightly higher rates than in the population. Overall, the fractions of households in the cohort(round 1+2) is notably lower in the cohort than in the Census for multiple-family residences,residences occupied by a single individual, residences with 0-1 vehicles and households with | t|Intercept 1.7245 0.1438 11.99


27/82

0

50

100

150

200

250

300

350

PM2.5

Em

ission

Rate(m

g/m

i)

regular participant

converted refusal

1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

Mod el Year

Figure 4. Aggregate PM emissions for Participants and Converted Refusals in Round 1 (Summer).

1000

PM2.5

Em

ission

Rate(mg

/m

i)

100

10

1

0.11965 1970 1975 1980 1985 1990 1995 2000 2005 2010

Mod el Year

regular participant

converted refusa l

Figure 5. Aggregated PM Emissions in Round 1 (logarithmic scale).

21


28/82

These tests failed to disprove an assumption of no significant differences in PM emissionsbetween the participants and conversions in Round 1. While not definitively resolving thequestion of the representativeness of the Round 1 measurements, these results allow a tentativeconclusion that non-response was motivated largely by reasons other than perceived highemissions. To the extent that this conclusion holds, it allows a further presumption that non-

response does not substantially affect the representativeness of the Round 1 results.

7.2 Emission Results

We present the results from the testing in Kansas City in several stages. First we briefly examinethe test-to-test variability observed by comparing the back-to-back measurements. Then weshow the general trends including: averages broken up by various testing regimes, averageemissions of smoking vehicles, and a comparison to past studies. Next we go into further detailon the temperature dependence of PM emissions. These temperature results are compared topast studies and also lay the groundwork for subsequent modeling. Then, we look at trends in

the aggregate LA92 emissions by model year and age as well as correlations with otherpollutants. In addition to furthering our understanding of PM formation from gasoline engines,this analysis will also help inform the development of new PM models for inventory analyses.Finally, in comparison to the aggregate results, we repeat some of the above analysis, but withthe separate bag by bag data from the LA92.

During both summer and winter phases, repeat tests were performed on selected vehicles, with15 repeats during round 1 and nine during round 2. In this case back-to-back means that testswere performed within several days of each other, not on the same day or at the same time. Ascan be seen in Table 7, the composite PM results show a high degree of variability for somevehicles, particularly during the winter. However, the direction and magnitude of differences do

not appear to show consistent patterns with regard to ambient temperature or humidity.

Absolute differences in composite results ranged from -2.9 to 5.6 mg/mi in summer and -183 to0.8 mg/mi in winter. Corresponding percent differences were -92 to 241% in summer and -97 to25% in winter. The most striking differences in absolute terms (-183 and -132 mg/mi) occurredin winter (duplicates 5 and 6). The temperature during both tests was between 35 and 45 F,showing that a large temperature differential does not appear to be responsible. Although notpresented here, the repeatability of repeat measurements for gaseous pollutants is substantiallyhigher than for particulate measurement. We present some hypotheses for extent of the test totest variability in a partner paper (Nam et al, in publication).

These results are valuable in that they give insight into the degree of measurement variability forindividual vehicles inherent in gravimetric measurement, which may be large enough to meritconsideration in analysis and use of the data.

22


29/82

Table 7. Composite Gravimetric Particulate Matter Results for Repeat Back-to-Back Tests during Rounds 1 and 2.

Round Duplicate Run # Temperature

(F)

Relative Humidity

(%)

Gravimetric PM

(mg/mi)

1

1

11

1

1

1

1

1

1

1

1

1

1

1

2

22

2

2

2

2

2

2

5

7

42

12

8

11

3

9

6

1

18

16

10

17

3

95

2

10

1

8

6

4

Test 1 Test 2

84060 84062

84104 84109

84110 8411584111 84116

84120 84123

84132 84137

84151 84156

84166 84169

84175 84180

84198 84200

84258 84262

84308 84312

84321 84328

84332 84341

84345 84350

84437 84442

84449 8445184465 84468

84482 84484

84485 84490

84537 84543

84541 84542

84627 84632

84690 84695

Test 1 Test 2

90.3 83.9

78.1 83

84.7 8488.4 87.6

93.6 85.1

80.3 76.9

69.4 65.4

71.5 70.7

70.1 74.6

65.8 65.2

70.8 71.8

77.4 74.3

80.5 82

76.8 76

74.1 70.7

60.1 40.7

25.8 37.637.9 36.6

39 40.7

38.9 36.9

40.9 49.5

49.8 44.6

47 45

55.2 42.4

Test 1 Test 2

63.8 80.3

55.7 59.1

56.8 57.151.6 49.5

44.4 56.2

38.3 34.9

46.8 59.4

39.2 44

47.8 44.4

63.2 68.8

1.5 50.2

42.3 61.1

33.3 40.7

39.7 54.3

42.5 49.5

47 59.6

39 61.356 46.3

70.1 56.8

59.2 55

63.8 52.5

68.3 61.3

34.9 30.4

38.3 67.6

Test 1 Test 2

0.994 1.236

1.556 1.069

10.006 7.5974.068 1.787

5.467 3.428

4.334 4.413

1.452 2.022

0.604 0.043

3.644 9.257

2.072 3.687

4.798 3.321

3.914 0.958

1.366 0.391

2.24 7.647

0.327 0.444

2.078 2.535

10.153 4.62188.706 5.223

14.104 8.658

20.047 3.842

3.178 3.982

6.332 4.908

232.116 99.412

2.005 2.119

A reporting of (certain) average emissions from a test program is more meaningful if the samplein the test program is corrected, or weighted, to the population of vehicles registered in thatregion. The fraction of vehicles of a particular model year recruited in this study may notnecessarily correspond to the fraction of vehicles of the same model year in the general fleet inthe region. Since the fleet figures are available to us, we present population weighted statisticswhere appropriate. These should be distinguished from VMT weighted rates, which may differsince older vehicles (while they exist) tend to be driven less, which would affect their relativeemissions impact in a region.

The population weighted average emissions are 12 mg/mi, though the distribution has a high

degree of skew. The maximum emission rate is 417 mg/mi from a 1973 truck measured at 62F.

The maximum emissions for a car is 260 mg/mi from a 1978 vehicle measured at 79F. The

average test temperature and (population weighted) emissions in the summer is 76F and 8

mg/mi, respectively, while in the winter it is 45F and 15 mg/mi.

The population weighted average emissions from cars is 11 mg/mi and for trucks it is 13 mg/mi.These results support the expected trends that trucks tend to have higher PM emissions than cars.There are two likely reasons why light truck emissions could be higher than those of passengercars. Trucks tend to have larger engine displacements and consume more fuel per mile.Additionally, if PM emissions loosely follow HC emissions, the standards for trucks have lagged

23


30/82

.275

.250

.225

.200

.175

.150

.125

.100

.075

.050

.025

0.000

.425

.32.350

.375.4005

.22.250

.275.3005

.07.100

.125.150

.175.2005

.025.050

0.000

those of cars. Also, emissions in the winter tend to be higher than they are in the summer. Wewill explore this in greater detail in a later section.

The following figures show the distribution of emissions in the summer and winter. In thisdataset, 296 cars and 182 trucks had viable PM data for a total of 478 tests (though some of these

vehicles had repeat tests). From a population weighted standpoint, 50% of the total emissions iscoming from 13% of the vehicles. Only 1 of the 14 highest emitters (defined here as emitting 50mg/mi or greater) in the summer was certified to Tier 1 standards (the rest were Tier 0), and ofthe 14, 6 were trucks while 8 were cars.

200160

140

120

100

100806040

Std. Dev = .0320

Mean = .014

N = 218.00 00

Std. Dev = .05

Mean = .025

N = 260.00

PM Composite (g/mi) PM Composite (g/mi)

Figure 6. Histogram of Gravimetric PM emissions composited over the LA92 cycle in the summer (left) and

winter (right).

Some of the vehicles were observed to have visibly smoking exhaust during idle. Of the 478vehicles tested, 406 were observed for smoke and categorized into four distinct smoke intensitylevels: No smoke (or normal), Low, Medium, and High (N, L, M, H respectively). We furthercategorize L, M and H as smokers and N as non-smokers. The smoking label has beenused in the past to indicate PM high emitters, though it is subjective definition. In the summerphase of testing, there were 10 (L only) smokers observed constituting 5% of the vehiclesmeasured for smoke. The population weighted average emission rate of these 10 vehicles is 43mg/mi. Of the highest emitting 14 vehicles in the summer, only 2 had observable smoke.

An earlier field study of smoking vehicles in California, found that between 1-2% of the lightduty vehicles in the fleet emit visible smoke [Durbin et al, 1999]. The average FTP emissions ofa select sample of smoking vehicles from this previous study, was 399 mg/mi. However, one

cannot compare the smoking rates of these two studies, since the vehicles in KC were notexplicitly recruited by their smoking classification. Moreover, a direct comparison of thefrequencies also cannot be compared. This is due to the fact that the observation methods werevery different between the studies. In KC the vehicles were observed for smoke while idling (atlow loads). In the California field study, smokers were observed when the engines were loaded:at signalized intersections and highway ramps. These vehicles may also have had more exhaustdissipation due to their higher speeds, and the root causes of smoke formation may be quite

24


31/82

different in idling versus the hot running engines. It is important to note that most analyses ofsmoking vehicles is more qualitative than quantitative.

In the winter phase, 40% of the vehicles were observed to be smoking to some degree. Of thehighest 26 vehicles tested (which contributed 50% of the winter emissions), 10 were smokers, 8

were not, and 8 were not observed. However, since it may have been difficult for the techniciansto distinguish definitively smoke from water condensation due to the cold temperatures, wintersmokers should not be grouped together with the rest of the population. Because of the highdegree of water condensation in the smoke, the frequency of smoking vehicles seen in this winterstudy is not representative of high PM smokers in the fleet.

These summer and winter results support the notion that not all smokers emit high PMemissions, and not all high emitters are smokers. However, smokers do tend to have higheremissions as demonstrated in the following table. These levels are considerably lower than theaverage smoker emission rates from past studies, which were recruited specifically for theirvisible smoke levels, thus potentially biasing their emission rates higher than this study.

Table 8 shows the population weighted average emission rate for each smoker category duringboth phases of the study.

Table 8: Population weighted average emissions by smoker classification.

SmokerCategory

PMmg/mi N

No 8.7 338

Low 19.7 68

Medium 49.0 21

High 63.8 8

We have discussed some of the issues with defining a PM high emitter. Choosing a cut-off valueis the primary challenge. Basing the cut point on statistics such as standard deviations, orcumulative distributions of emissions is overly arbitrary and dependent on the sample. On theother hand, several previous authors employ smoker as a surrogate definition. We have shownthat though there is a correlation with higher PM emissions and smoking vehicles, thatrelationship is subjective and can even be misleading at times, especially in colder temperatures.Finally, a definition based on an emission standard (as with HC, CO and NOx) is also notfeasible because PM standards for gasoline fueled vehicles were chosen to mirror those fromlight-duty diesel. Thus they were not chosen at a stringent (technology forcing) level. Thefollowing figure shows the scatter plot of PM emissions compared to the emissions standard.Clearly, the standard is not a good measure of high PM emissions, especially since there were no

standards before 1996, and the test program uses a different drive schedule (and temperature)than the certification test.

25


32/82

100

50

0

1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

Model Year

150

200

250

300

Cars

Truck

PM10 Std

Figure 7. Scatter plot of KC PM emissions as a function of model year, compared to the PM10 emissions

standard.

Unfortunately, this study does not answer the question of how to define a PM high emitter. Theabove analysis describes some characteristics of PM high emissions without necessarilyadding a high emitter label or cut point. In a later section of the paper, we show cumulativedistributions with temperature adjusted data. However, this will have to remain a subject offurther study.

7.3 Comparison with past studies

Table 10 compares the results of the present study with two major studies that have preceded it:CRC E24 phases 1, 2, and 3, as well as the Gas Diesel Split (GDS) study. The emissions ratesare averaged by the model year strata as used in the Kansas City study to provide a commonbasis of comparison (though a more recent model year group was added at 2001+). Wherepossible, smokers that were specifically recruited as part of prior studies are omitted from thecomparison, since their inclusion may lead to over-representation of high emission vehicles. Thesmokers included in the normal recruitment (on the basis of model year or mileage) wereincluded in the following analysis. A comparison of these past studies with the present one isinformative, but it is important to note that all of the studies are very different, thus making

comparisons qualitative in nature. Mainly, a model year comparison is limited since the samemodel year vehicles in KC would be approximately 8 years older than those in E24, and 3 or 4years older than those in the GDS, and thus probably have more miles driven. Moreover, thenumber of vehicles recruited in each of these other studies was smaller, thus making estimates ofaverages more unstable.

As mentioned earlier, the CRC E-24 project was conducted in 3 phases. Phase 1 is morecommonly known as NFRAQS. Table 9 describes the differences between the programs:

450

400

350

PMEmissions(mg/mi)

26


33/82

Table 9: A comparison of 5 different PM measurement programs.

CRC E24-1 CRC E24-2 CRC E24-3

Gasoline-Diesel

Split Kansas City

Region

Northern Front

Range, CO SCAQMC

Dallas-Fort Worth,

Houston Los Angeles Kansas City

Recruitment

Local Merchants,

etc.

Registration +

phone

SwRI employee

vehicles + local

school

BAR via smog

check + local

advertising

Random digitdialing +

registration

database

Test Year 1996-1997 1996-1997 1996-1997 2001 2004-2005

Drive Schedule FTP FTP FTP LA92 LA92

Temperatures 23-67F in winter Lab Lab 64-99F 20-100F

PM size PM10 PM10 PM10, PM2.5 PM2.5 PM2.5

Fuel Local Local Local Local Local

Fleet Certif icatio National CA National CA National

Given the time span between the different studies, it is unexpected that the emissions levels areapproximately consistent (within 95% confidence bounds). The exception is in E24-2, where the

vehicles have significantly lower emissions in two of the model year groups. This unexpectedconsistency across the studies could be due to a number of factors: the KC fleet within thesemodel year groups may be cleaner than the other studies, the other studies may haveoversampled dirtier vehicles, a survivor effect may be keeping the oldest emissions rates fromexceeding a certain level by a process where owners scrap the dirtiest vehicles in the fleet, or theemission rates may be mostly dependent on model year (or technology), i.e. emissions rates maynot deteriorate significantly over time. The latter is an unlikely hypothesis. It is likely, however,that a combination of these factors is in play.

Table 10: A comparison of average emissions by model year groups from 5 different test programs: KC, CRC

E24 phases 1, 2, 3, and Gasoline Diesel Split Study. PM ttl has no PM size preselection.

Model Yr E24_1_Summer E24_1_Winter E24_2 E24_3 Gas-Diesel Split

Group KC Summer 2004 KC Winter 2005 1996 1997 1996 - 1997 1996 - 1997 2001

PM2.5 N PM2.5 N NPM ttl NPM ttl NPM ttl NPM ttl NPM2.5

pre-1981 69.0 58.3 8 75.7 40.6 24 95.4 34.7 25 78.3 36.3 15 34.0 12.2 14 148.5 94.4 10 59.0 54.4 6

81to90 29.8 12.7 56 36.4 11.8 68 46.0 24.1 47 37.8 11.4 33 28.1 19.1 54 64.4 29.3 25 34.2 16.2 18

91to95 9.7 3.6 48 20.0 6.9 72 2.5 0.7 17 30.4 32.6 7 2.8 1.2 50 5.8 1.2 12 3.6 1.6 26

96to2000 3.7 1.4 77 8.9 3.1 61 4.5 5.0 3 3.0 1 2.3 2.2 11 6.4 1.7 6 3.9 3.7 3

2001+ 2.2 0.6 29 6.8 2.7 35

In the following sections, we look at trends of PM as a function of a variety of independentvariables. Such variables include temperature, model year, and age. We also examine thecorrelations with other pollutants since the mechanisms of formation of the more traditionalcriteria pollutants (HC and CO) are better understood. It follows that these trends may shed lighton PM formation processes as well. The correlations may also show how controls of HC may ormay not influence PM emissions.

27


34/82

7.4 Temperature Effects on Composite Data

In this section, we examine the effects of temperature on the KC data. The vehicles tested inKansas City were tested under ambient temperature conditions. Before other PM trends can be

determined, we must first isolate (or adjust for) the effect of temperature. Some studies in thepast (cited below) have shown that PM emissions increase as temperature decreases. However,this effect has never been conclusively quantified and compared across studies. This is what wehope to accomplish.

The first pass at the analysis is conducted on the composite (or aggregate) data only. Compositedata consists of the 3 bags of LA92 emissions data weighted by the appropriate weightingfactors. Temperature effects are scrutinized in 3 ways: by examining all the data, the correlationvehicle, and the summer-winter paired data. The correlation vehicle (1988 Ford Taurus) wasmeasured 24 times throughout both phases of the program, each test occurring at a distinctambient temperature. Also, 43 vehicles were tested, both in the summer and winter. This paired

test database is valuable in looking at the fraction of the emission rate variability explained bytemperature. The paired test data is shown in Figure 8. The figure clearly shows that winteremissions exceed summer emissions in nearly all cases. Some vehicles have 10 times more PMemissions in the winter as in the summer. This leads us to conduct much of our subsequentanalysis in log-space, which allows for quantification of effects across a large range of results.

10 times higher

in winter

Figure 8. Scatter plot of winter vs summer PM emissions on log scale (EPA, 2008).

28


35/82

The PM trend with temperature is shown in Figure 9. The red (empty) diamonds are all thevehicles (except the correlation vehicle) and a trend line is drawn through this grouping. Thesolid blue diamonds are the summer-winter paired tests. And the solid blue squares are thecorrelation vehicle repeat tests. The slope for all the data is -0.0188 (in log space), and for the

correlation vehicle is -0.0415.

lnPM

7

6

5

4

3

2

1

0

-1

-2

y = -0.0188x + 3.0474

R2= 0.0645

y = -0.0415x + 3.6049R

2= 0.6632

20 30 40 50 60 70 80 90 100

Temperature

all vehicles

Correlation vehicle

Paired tests

Linear (all vehicles)

Linear (Correlation vehicle)

Figure 9. lnPM vs test (ambient) temperature for all vehicles, paired tests, and correlation vehicle.

The matching paired tests each have separate slopes which were calculated using the followinglinear equation in log-space.

m = (lnPM2 lnPM1)/(T2-T1)

Pairs were omitted from the analysis if any of the conditions exist: one of the PM values is

missing; the temperature difference between two tests was less than 10F. If the temperature

difference is less than 10F, the test-to-test variability dominates over any temperature effectsand the slopes become ill-defined. Unfortunately, these criteria eliminate 10 of the 43 pairedtests. The remaining slopes are plotted by model year in Figure 10. The average of the slopes is-0.036 +/- 0.009 (95%CI) (we will explore a more robust method for estimating this slopeparameter next). However, the plot seems to indicate that there is no apparent model year trendwith temperature. This leads us to believe that temperature effects on PM emissions areindependent of vehicle technology. Also, though not shown, the trend does not appear to be

dependent on average temperature of the repeat tests.

29


36/82

SlopeoflnPMvsTem

-0.1

-0.09

-0.08

-0.07

-0.06

-0.05

-0.04

-0.03

-0.02

-0.01

0

Delta TemperatureSummer-Winter Paired Tests

avg = -0.036+/- 0.009

1973

1979

1987

1989

1989

1993

1994

1995

1995

1997

1997

1998

1998

1999

1999

2000

2002

Figure 10. Temperature slope from individual matched pairs as a function of vehicle model year.

In order to estimate the effect of temperature from the summer/winter matched pair vehicles, werely on a univariate general linear model, run on the SPSS statistical software. The vehiclewas treated as a fixed factor (categorical variable) and the correlation vehicle was weighted by afactor of 0.09 in order to give it the same weighting as the other 33 matched pairs. Otherwise,the 22 points of the correlation vehicle would dominate the linear model. This is the temperatureeffect that should be used in order to adjust all of the PM data to a baseline temperature forsubsequent comparisons of summer and winter data. The matched vehicle slopes are showngraphically in Figure 11. There are clearly some vehicles with a weaker temperature effect and

others with a stronger one, however a statistical mean can be discerned. Using the solution to theprevious equation, the correction will be applied in the following manner:

lnPM2 = -0.03356*(72F T1) + lnPM1

This is based on the assumption that it is the slope that drives the temperature effect; thus theoffset in the slope is defined by each individual test and is ignored for the purposes of thistemperature correction model.

30


37/82

0.1

1

10

100

1000

10 100

PM(mg/mi)

20 30 40 50 60 70 80 90

Temperature

Figure 11. 33 Matched pair vehicles as a function of temperature on log scale.

Temperature effects on PM emissions have been evaluated in some recent testing programs. In

2000, Southwest Research Institute measured the PM from 7 vehicles at 30F and 75F[Whitney, 2000]. In 2001, the EPA (at the Office of Research and Development) conducted

testing on 9 vehicles with model years ranging from 1987 to 2001 at 75F, 20F, 0F and some at

-20F. This is referred to as the ORD data [Stump, et al, 2005]. In 2005, EPA tested 4 Tier 2

vehicles at 75F, 20F and 0F in support of the Mobile Source Air Toxics (MSAT) rulemaking[Stanard, (2005)]. The results of these three test programs are compared with the Kansas Citydata on Figure 12. Note that the trends (slopes) are quite similar, and that the slope for the KCdata lies in between the studies . The MSAT data is considerably lower, since the programspecifically targeted Tier 2 vehicles. It is again, interesting to note that temperature trends(slopes) seem to be independent of vehicle technology, since the slopes are relatively constantfrom the various test programs.

31


38/82

y = -0.0353x - 4.2332

R2= 0.7284

y = -0.0385x + 3.7822

R

2

= 0.8935

y = -0.0258x + 3.9758

R2= 0.2722

-10

-8

-6

-4

-2

0

2

4

6

-20 0 20 40 60 80 100

Temperature

lnPM

SwRi Test vehicles

ORD Test vehicles

KC Correlation vehicle

KC SW Matched pairs

MSAT Test vehicles

Linear (KC vehicles)

Linear (MSAT Test vehicles)

Linear (ORD Test vehicles)

Linear (SwRi Test vehicles)

y = -0.0336x + 3.33

Figure 12. PM temperature trends from 3 different test programs in log space.

Figure 13 shows the some of the same data as the previous figure but in terms of absolute PMemissions. The fit is to the KC data, which is consistent with the ORD data despite the verydifferent fleet mix tested.

0

20

40

60

80

100

120

140

160

PM(mg/mi)

KC SW Matched pairsfit to KC w/ correction

ORD Test vehicles

-20 0 20 40 60 80 100

Temperature

Figure 13. PM temperature trends from 3 different test programs in linear space. The fit the KC data is MSE

corrected.

32


39/82

A similar analysis of temperature trends for the other pollutants was conducted and the resultsare shown in Table 11. All of the slopes are significantly different from zero at the 5% level.PM is the most sensitive to temperature, whereas for some other pollutants the effects are small.For the remainder of this report, the correlation vehicle is usually omitted from all trendreporting since the vehicle is not part of the randomly sampled population, and the emissions

from this vehicle is not necessarily representative of what would be expected to be seen in theexisting fleet.

Table 11. Exponential emissions dependence on temperature (P


40/82

-1.25

-1.75

-2.25

-2.75

-3.25

-3.75

-4.25

-4.75

-5.25

-5.75

-6.25

-6.75

-7.25

-7.75

-8.25

-8.75

-1.25

-1.75

-2.25

-2.75

-3.25

-3.75

-4.25

-4.75

-5.25

-5.75

-6.25

-6.75

-7.25

20 30

20

10

10

Std. Dev = 1.39 Std. Dev = 1.28

Mean = -5.40 Mean = -4.58

N = 218.00 0 N = 260.000

LN(PM in g/mi) LN (PM in g/mi)

Figure 14. histogram of the logarithm of summer (left) and winter PM emissions (right)

Figure 15 shows cumulative frequency distributions of composite emissions, with rounds 1 and 2

combined by (model year) stratum. Note that the values from round 2 have been adjusted fortemperature using the equation in the previous section. However the distributions have not beenpopulation weighted, so the distributions are not necessarily indicative of the actual KC fleet.

The distributions show expected patterns, with the oldest vehicles showing highest emissions, theyoungest vehicles the lowest. In addition, within each model-year group, values for cars andtrucks are distributed more closely than are values for adjacent model-year groups. In absoluteterms, the distributions show that the skew in the data is more pronounced for older vs. youngervehicles. For example, the difference between the 90thand 50thpercentiles is approximately 4.0,6.5, 60 and 100 mg/mi for cars in the 96+, 91-95, 81-90 and pre-80 groups, respectively. Theskew in the distributions is also shown by the differences between medians and means, which

fall roughly between the 70

th

and 80

th

percentiles.

Figure 16 shows the same cumulative frequency distribution as Figure 15 but on a logarithmscale. One can note that the distributions look nearly log-normal, though in some of the stratum,the high emissions tail appears to be slightly longer than the low side.

34


41/82

Figure 15. Cumulative distribution of PM emissions (temperature adjusted to 72F) for the model year

groups.

Figure 16. Cumulative distribution of PM emissions (temperature adjusted to 72F) for the model year groups

on a logarithm scale.

35

8/13/2019 Analysis of Particulate Matter Emissions Fro

analysis of particulate matter emissions from light-duty gasoline vehicles in kansas city

Documents